Research On Virtual Network Mapping Algorithms

In the new century, the number of Internet business just explodes and the characteristic of business diversification becomes more and more obvious. This raises new challenges to the traditional Internet architecture, which has an increasing inability to meet the growing business requirement. However, It will be politically and financially unfeasible to rebuild a brand-new replacement of the traditional Internet architecture. The network virtualization technology emerges in this situation.At present, the most difficult technical challenge of the netwrok virtulization is virtual network mapping(VNM), which aims at maximum utilization of the physical network resources in meeting the request of network virtualization.This thesisfirstly makes the research onthenetwork virtualization technology and VNM problem and summarizes the heuristic algorithm which is widely used in solving the VNM problem. The performance of some algorithms covered in the thesis, Genetic Algorithm(GA), Ant Colony Algorithm(ACA), and Greedy Algorithm, is also simulated and analyzed. Based on the research and analysis, a new mapping algorithm is proposed combining the advantages of GA and ACA. The proposed algorithm introduces the positive feedback mechanism of ACA into the mutation operation of GA, thus getting rid of the blindness of mutation operation and improving the performance of algorithm.To verify the correctness and advantages of the proposed algorithm, a medium-scale virtual network mapping model based on the For CES(Forwarding and Control Elements Seperation) network is constructed, aiming at the maximum acceptance ratio of virtual network requests and the maximum utilization of physical network resources. Genetic Algorithm(GA), Ant Colony Algorithm(ACA), and Greedy Algorithm are used for the resolution, respectively. The performance simulationresults shows that the proposed algorithm is feasible and efficient in solving the VNM problem.